As I was reading Emergence, one particular passage grabbed my attention. On page 127, Johnson is drawing comparisons between the Internet as an emergent system and the human brain. He describes how new software is scanning the connections in the surfing behaviors of internet users. These methods depend not on the actual content being read, but on the complex relationships between different sites. He then compares this to how the human brain functions — since the firing of neurons is relatively slow compared to the circuitry of a computer, our brains don’t think about things linearly – instead, it’s an enormously parallel machine that looks for underlying patterns.
If millions of different pieces of software do this pattern matching just like human neurons do, how does that change our view of the Internet as, perhaps, a emergent entity? I believe that one of the important differences between that pattern-matching is that there is no real connection between the software that scours the internet. The greatest strength of our brains is that the overheads of parallel processing – the time it takes to communicate, how to pass information, what to do with idle processors (neurons) – are nearly invisible, simply through the the millions of years of evolution. On the other hand, the software we have is created by humans. Communication and what to do with the data must be guided and filtered by human hands – and the sheer complexity of the task is no small thing. These multitudes of programs be truly emergent until real, unforeseen behaviors arise. Perhaps it can learn and guess at which websites will be popular.
As software that crawls the internet grows more complex, I think we’ll see more interesting trends emerge from the woodwork. The software that deals with interconnections between websites effectively have the ability to predict whether a new website will be popular or not – and beyond that, they have the necessary information to perhaps create a new, popular website. If this software begins to create it’s own internet pages…well, that may be a spark of something interesting.